The paper addresses the effect of a large wind farm cluster on the frequency and magnitude of sea spray freezing events. The conclusion is that the effect of even a very large cluster, and in an environment prone to these events, is minimal. The paper thus ends the scientific debate on this topic.

There is potential for improvement. Inconsistency and inaccuracy in how the authors present the hypotheses, scope, methods, and results, are plentiful and should be corrected. For example, it has evaded the senior authors' detection how "freezing" and "icing" are used interchangeably due to apparent lack of understanding of the two separate phenomena. Please see the specific comments below.

Given the paper's light scientific weight, it may be published at the editor's discretion, after these comments are taken into account and the paper revised accordingly. Therefore the required revision is classified as "minor". 

Abstract:

15-16, 21 "liquid precipitation" is not the most common cause of ice accretion. Liquid cloud particles cause most of the icing which causes the aerodynamic degradation of the wind turbine blades, which in turn causes "extra load and fatigue" and loss of production (not mentioned in the Abstract - please add). Icing on the blades is usually not called "freezing" and I suspect that the authors refer to for example freezing of the sea spray on the service boats and access platforms. This inconsistency repeats throughout the paper - please fix it.

24 "not all freezing events are cold enough to signify a cold air outbreak" ... unclear what the meaning of this statement is. Cold air outbreak is mentioned several times in the paper, unnecessarily.

26 The wakes are said to "mitigate the chance for freezing". Given that the effect of wakes to wind farm performance is significant and more important than freezing, this statement should perhaps be revised.

 Introduction:

34 Is the White House a credible scientific source/reference?

39 The effect of ice on energy production is mentioned here. Please include it into the Abstract (see my comment above)

42 If the rotation stops entirely, then one would expect the power production to reduce by 100%, not just up to 80%.

43, 45 The reference is irrelevant, why are just two case studies selected? Please consider together with the references in 49 (Martini et al. ...). The turbine blade icing effect is indeed well studied, and many of the 542 references in the IEA Task 19 technical report (https://iea-wind.org/wp-content/uploads/2021/09/Lehtomaki-et-al.-2018-Available-Technologies-for-Wind-Energy-in-Cold-Climates-report-2-nd-edition-2018.pdf) would be more appropriate. Generally, the reference should point to the earliest appropriate publication, not to a random one.

47 Is there evidence that the winds are faster in cold air outbreaks, than in for example warm air outbreaks?

69 Minor note: turbulence does not transport temperature, but heat.

77 It is perhaps interesting, but not "crucial" to understand how large scale deployment of wind farms will modify freezing events. 

82 What is "post-production effect" in this context?

Methods:

157-161 Spray freezing and riming are two distinct phenomena and the paragraph is not sufficiently clearly introducing them as such.

188 Do the convective rolls have any meaning in the context of sea spray freezing, or are they just relevant for the in-cloud icing? Please clarify.

194 Is perhaps the temperature at 2 meters meant here, and not at 10 meters? There are more occurrences of the 10 m temperature. Please check.

Results:

221-226 Is this current dynamics supposed to help explaining the results regarding the freezing events. If yes, then proof is required, otherwise it is just speculation. If no, then it is not necessary.

229-230, Figure 2 caption. Please normalize the color scales on the two plots so that they can be compared. For example, show the number of hours per season.

234 "freezing conditions" is slightly vague, especially since you calculate the magnitude as well. Could you perhaps use the magnitude even more?

236 Again, more than the area (12 times the wind plants), the severity of the freezing events would be more important to discuss here.

245 Figure 3. It is not immediately clear where the zoom fits. Please consider redrawing.

260 "253 hours" seems inconsistent with the total which is  182 (or 187). Please check or clarify.

261 "light ice" here helps to slightly resolve my comment about the freezing severity, above.

265 It would be more appropriate to express the pressure gradient in hPa per 100 km (the value 4 would then mean 4 times the geostrophic wind speed of 10 m/s) - just a suggestion.

266 The reference is weird. The geostrophic wind and how it is calculated was first mentioned in 1857. In meteorology, work of e.g. Bjerknes would also be a meaningful reference. Again, the references should point to the earliest appropriate publication, not to a random one.

281 The meaning of CAO in the context of freezing is not clear. Does it matter if an event is called CAO? Especially since one uses the same variables to calculate if an event is CAO, and if there is freezing.

287-288 True statement, the grid efficiency does suffer from high temperature, but is irrelevant in the context of freezing. Please consider removing, or explain why this is important for this paper.

309 The total effect of up to -0.041 K is so small that it should perhaps be pointed out even more, how small the effect of wind farms to the freezing is.

315 Figure 5. Which height above the surface is this?

320 Icing or freezing, blades or ship? Please clarify.

322 Here you say freezing. It is really not OK to use icing and freezing interchangeably like this!

337 13 hours, compared to what? Please express as fraction.

345-346 "... flow acceleration is present ...". "may be present" would be more accurate, it is not relevant for his paper, so why mention it.

347-348 The statement about the numerical noise seems to negate the rest of the analysis. The physics of freezing is correctly captured in the models, and the results are consistent. It is true that WFP can introduce noise, even at the opposite side of the planet in e.g. MPAS model. Please provide more results supporting the numerical noise hypothesis, or consider removing the statement.

Dear authors,

The article is well written and interesting to read. The subject is original within the field of wind energy research and is of relevance for the future offshore wind farm development of the east coast in the USA. The results show that there is considerable icing risk in the mid-Atlantic offshore wind farm areas and that the effect of wind farm wakes on icing risk is minimal. The manuscript would benefit by considering the points below.

Specific comments:

Regarding the two sentences and corresponding references:

“Some observations indicate that excessive icing can reduce torque enough that blade rotation stops entirely, causing up to 80 % reduced power production for a single turbine”.

“Some turbines have icing detection and mitigation technology included at added cost, although current strategies need improvement (Madi et al., 2019).”

I believe these statements are a bit outdated, especially seen in the light that the paper is mainly focusing on future wind energy scenarios. I do not want to make an advertisement for specific solutions, but there are many de-icing, or anti-icing solutions where power reduction can be avoided (see e.g. https://www.iqpc.com/media/1001147/37957.pdf, https://wicetec.com, https://www.video.vestas.com/video/21313125/vestas-anti-icing-system). I suggest that the text should be updated to state that it will be important to include some proper anti-icing solutions.

It becomes clear after reading section 2.1 that the NOW-23 data set is also based on WRF runs. I would suggest mentioning that at the beginning of section 2.1.

Table 1 is a bit strange. Maybe there should be a column with simulation type number 1-3, or something similar that could be referred to in the text. “Turbine type” should be “turbine rated power”, or if you want to keep turbine type then mention the type. The period does not need a column (as it is the same for all simulations) could be mentioned in the figure caption.

Line 152: Why can SST be replaced by skin temperature?

Have the authors investigated how the results would change if you would use the most conservative thresholds? It would be beneficial to include a sensitivity study about that, e.g. on a small subset of data or at the POI.

Why is the acronym for predictability chosen as PPR?

Fig. 2: It’s a little bit confusing that the turbine locations are shown for these simulations that were performed without turbines. Could you state in the caption that the locations are shown for illustrative purposes, but were not included in the simulation results?

Fig. 7: Is it percentage or difference in whole hours? If percentage, the color bar label needs to state that by adding e.g.  [%]. Could you include smaller intervals on the color scale, so it’s possible to see more variation?